The new edition of the APA Monitor magazine has an article that discusses the psychological impact of solitary confinement in light of its growing use in American prisons.
One of the most interesting points is that evidence for the effect on solitary confinement on prisoners is actually quite limited due to difficulties studying incarcerated people.
Prisoners in administrative segregation are placed into isolation units for months or years. Corrections officials first turned to this strategy in response to growing gang violence inside prisons, Dvoskin says. Though critics contend that administrative segregation has never been proven to make prisons safer, use of this type of confinement has continued to rise. That’s worrisome to most psychologists who study the issue. Deprived of normal human interaction, many segregated prisoners reportedly suffer from mental health problems including anxiety, panic, insomnia, paranoia, aggression and depression, Haney says (Crime and Delinquency, 2003)…
However, much of the evidence of harm comes from cross-sectional studies or research done on people who are not in prison, such as the isolated elderly. Designing a long-term study to follow prisoners in solitary confinement is challenging. Each correctional system is unique, inmates move in and out of segregation, and many states prohibit or limit psychological studies of incarcerated individuals due to ethical concerns.
Most of our evidence, it turns out, comes from other people deprived of human contact, although the effects have been found to almost universally unpleasant until now.
Link to article ‘Alone, in the hole’ (via @ResearchDigest)
Mind Hacks 
I don’t think it would be impossible to design a study that would, at the very least, hit the high points of the effects of solitary confinement. The general public might be surprised at the complexity (which, in other words, means that they would extinguish any examination of the issue)of creating such a study, though it might be accomplished retrospectively even now.
I began to wonder if solitary was a resource-consuming idea that could be replaced by some other strategy. I would imagine that solitary might work, short-term, for some inmates; but wouldn’t transferring two- or three-time offenders of basic rules (no acting-out, no stealing, no violence, etc.) so that they would have to recreate a new social structure each time they moved? That seems more powerful to me. The more you make trouble, the more times you have to reintegrate your social system. While disruptive, it seems more humane (but only studies would show if or how, and studies aren’t much wanted in institutions that don’t really care about their residents).
Also, presumably, the contact involved in a study would undermine the toture?